update broke gateway

[gritty gyro]

Updated openclaw. Gateway stopped working. Fires bonjour plugin followed by several channels failing then shuts down immediately after. tried removing bounjour and hard resetting gateway and no fix yet - same issue occurring. Openclaw gateway status shows: "Runtime: stopped (state Ready, last run 1, last run time 4/27/2026 9:41:36 PM, Task Last Run Result=1; treating as not running.)
Connectivity probe: failed
Probe target: ws://127.0.0.1:redacted
connect ECONNREFUSED 127.0.0.1:redacted
Capability: unknown

Service is loaded but not running (likely exited immediately).
File logs: redacted
Logs: schtasks /Query /TN "OpenClaw Gateway" /V /FO LIST
Restart attempts: redacted"

[rough badger]

That ECONNREFUSED is downstream — it means the Gateway already exited. Since this is Windows/schtasks, I’d separate service wrapper problem from Gateway crash first; Windows scheduled-task starts/restarts have had a few known rough edges. (github.com)

Please paste:

openclaw --version
node --version
openclaw gateway status --deep
openclaw logs --limit 500 --local-time

Then try a foreground diagnostic run:

openclaw gateway stop

$env:OPENCLAW_DISABLE_BONJOUR = "1"
$env:OPENCLAW_SKIP_CHANNELS = "1"
$env:OPENCLAW_LOG_LEVEL = "debug"

openclaw gateway run --verbose --ws-log compact

Interpretation:

• If this stays running: core Gateway is okay; the crash is probably channel startup, not Bonjour.
• If this still exits: we need the first fatal stack/error after startup — likely config/runtime/service issue.
• If it only fails when using gateway start/restart: likely Windows schtasks wrapper issue; foreground gateway run confirms that.

Bonjour is safe to disable for isolation; it only controls LAN discovery advertising, not whether the Gateway should run. (docs.openclaw.ai)

If this looks like a regression after update, run:

openclaw gateway diagnostics export --output openclaw-diagnostics.zip

and file it on GitHub with the zip + version. The docs recommend status, gateway status, logs, doctor, and diagnostics export for this exact class of “service installed but process does not stay up” issue. (docs.openclaw.ai)

[gritty gyro]

got up to verbose gateway mode - gateway persisted, i shifted over to trouble shooting with my agent, provided the context of this conversation so far, they commited disable of bonjour as a change to start then suggested closing diagnostic foreground and running gateway normally, that failed, i attempt foreground diagnostic again and now this is also broken: "22:59:45 [openclaw] Unhandled promise rejection: CIAO PROBING CANCELLED
22:59:45 [openclaw] wrote stability bundle: C:\Users*redacted*.openclaw\logs\stability\openclaw-stability-2026-04-27T12-59-45-639Z-96200-unhandled_rejection.json"

[rough badger]

Yep, that’s a real Bonjour/ciao crash signature — not “channels failing” anymore. The changelog has an explicit fix for CIAO PROBING CANCELLED, so first check your exact version. (raw.githubusercontent.com)

Run this:

openclaw --version
openclaw config set discovery.mdns.mode off
openclaw plugins disable bonjour
openclaw config get discovery.mdns.mode

Then test foreground again with env set in the same PowerShell window:

openclaw gateway stop
$env:OPENCLAW_DISABLE_BONJOUR = "1"

$env:OPENCLAW_SKIP_CHANNELS = "1"
$env:OPENCLAW_LOG_LEVEL = "debug"
openclaw gateway run --verbose --ws-log compact

Important bit: $env:... only affects that foreground shell. If “normal gateway” means the Windows Scheduled Task, persist only Bonjour disable like this:

[Environment]::SetEnvironmentVariable("OPENCLAW_DISABLE_BONJOUR", "1", "User")
openclaw gateway restart

OPENCLAW_DISABLE_BONJOUR=1 is the documented mDNS/Bonjour kill switch, and discovery.mdns.mode: "off" is the config-level equivalent. (docs.openclaw.ai)

If it still touches ciao/Bonjour after all three are true, that’s very likely a stale/broken packaged plugin/runtime path. Please attach the stability bundle it wrote plus: